A consideration for changing our PET data saving practices: a cost/benefit analysis

Objectives Every PET scanner produces raw acquisition data which is used to generate an image, and this data is almost universally discarded beyond its initial use. Most scanners do not have the capacity to store raw data, and the few that do require expensive license options. Our current PET protocols, particularly our data saving practices, stem from a time in which data storage was significantly expensive. However we can recognize that computing capacities have undergone significant changes. Data storage has become cheaper by orders of magnitude, and processing capacities have given us tools to innovate in new directions. Access to raw PET data has already been shown to provide a foundation for several areas of research and potential clinical applications. In this work we review our current data saving and utilization practices and consider how updating them may affect practice and innovation.

Methods The costs and benefits associated with changing data saving practices to make data more ubiquitously available were reviewed. Costs were determined monetarily, adjusted for inflation, and related to the total cost of a clinical PET scan. Potential benefits easily accessible PET data were also reviewed. The barriers to changing practice were also considered.

Results Costs for saving raw data were very high when PET was originally developed, but have been exponentially decreasing since. In 1990, the cost of archiving a 2Gb raw PET data file was prohibitive, costing approximately 330% the price of a clinical PET scan (assumed to be $5000 in 2015). In 2000 saving data became much cheaper, costing only 0.35% of the total scan charge. However it still lacked the required justification. As we enter 2016, we can recognize that the cost of saving raw data has plummeted into the negligible region, approximately 0.0005% the cost of a PET scan. At the same time we can identify areas of data driven research that may benefit from changes to practice that make raw PET data available. These topics include image reconstruction, dose optimization, motion correction (respiratory, cardiac, head), and cardiac flow. These are all areas of study that have utility in having access to raw data, and have been established in existing literature. The barriers to changing our practice to saving raw data include: (1) the lack of mainstream interest to make this change, (2) lack of standardized data format, and (3) the present proprietary nature of the PET listmode data. To date there has been virtually no discussion as to who rightfully owns the data and should have access to it (i.e. the patient, the hospital, or the vendor). The current status quo is that data utilization remains in the auspices of the vendor control.

Conclusions In PET imaging patients are exposed to ionizing radiation and it is incumbent upon us as a community to ensure that their data is being utilized to maximum benefit. The digital aspect of PET is supported by an electronic landscape that has undergone significant transformation. The cost of saving raw PET data has become virtually negligible. The benefits of changing our data saving practices are not yet established, but the concept of data supporting innovation has been established existing proof of principle research. While considering potential we can find precedent of access to data preceding academic and commercial innovation, as this is what took place when DICOM image format was established. Because every PET scanner generates raw data, simply updating our data saving practices can enable a new low cost, ubiquitously available resource that can support research and clinical innovation. Changing our data saving practices would likely require a concerted effort between users, vendors, and field leadership. Looking forward, we can expect technology will only bring us even more reductions in storage cost and faster processing, it is likely not an “if” but a “when” we change or data saving practices. We urge the field to be proactive in ushering in this transition. $$graphic_48B17598-1EB5-47A9-B8A7-ED2858A2F19F$$